-
Notifications
You must be signed in to change notification settings - Fork 9
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Merge TiKV near 7.4 #351
Merge TiKV near 7.4 #351
Conversation
) close tikv#15109 This PR updates the way to calculate the last_change_info, which is not very proper before. This also fixes a assertion failure in `next_last_change_info` that might happen when it executes in parallel with GC in compaction filter. Signed-off-by: MyonKeminta <[email protected]>
ref tikv#14900 Signed-off-by: husharp <[email protected]>
close tikv#15107 Signed-off-by: husharp <[email protected]>
close tikv#15122 Signed-off-by: hillium <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com>
…5185) ref tikv#15108 * raftstore-v2: support unsafe recovery create peer Reuse split mechanism to create peer. Because in v2 the only way to create and initialize to peer is via a snapshot message. * raftstore-v2: support unsafe recovery destroy peer Signed-off-by: Neil Shen <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com>
Signed-off-by: lijie <[email protected]>
ref tikv#15108 * raftstore-v2: support unsafe recovery demote * raftstore-v2: do not step force leader on confchange Signed-off-by: Neil Shen <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com> Co-authored-by: Connor <[email protected]>
ref tikv#15273 Signed-off-by: Neil Shen <[email protected]>
…ikv#15284) close tikv#15282 trigger compaction when no valid split key can be found Signed-off-by: SpadeA-Tang <[email protected]>
…kv#15249) ref tikv#15192 Signed-off-by: 3pointer <[email protected]> Co-authored-by: tonyxuqqi <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com>
…it. (tikv#15247) close tikv#15246 Modify the return value of the scan function under the RawStorage trait. * fix the scan function only return Value. Signed-off-by: joker53-1 <[email protected]> Co-authored-by: Andy Lok <[email protected]>
ref tikv#12842 - Restore merge status when there's an error between trim check and propose. - Fix a case where snapshot just before prepare merge can cause rollback. - Mute an assertion that fails in v1 bulk_insert testing. - Use SkipUntil in compaction filter, together with the newly introduced UnsafeFilter API, this should remove any tombstones regardless of the context. Signed-off-by: tabokie <[email protected]>
ref tikv#15258 Signed-off-by: Neil Shen <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com>
close tikv#15288 Add error type for "mismatch peer id". Signed-off-by: you06 <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com>
close tikv#15310 Rust uses CamelCase for struct names. And add a test for #[doc(hidden)] field. Signed-off-by: Neil Shen <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com>
close tikv#15200 reject read index if the applied index + raft_log_gc_count_limit <= read_index Signed-off-by: bufferflies <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com>
close tikv#15296 Signed-off-by: hillium <[email protected]> Co-authored-by: qupeng <[email protected]>
ref tikv#15276 Signed-off-by: hillium <[email protected]> Co-authored-by: qupeng <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com>
ref tikv#15056 Add logs for split region so that we know the time a split takes end-to-end: from the split proposed to the last splitt region finishes the initialization. Signed-off-by: tonyxuqqi <[email protected]>
close tikv#15324 Ribbon filter may cause OOM when compacting a sst with many entries. See tikv#15324 Signed-off-by: Neil Shen <[email protected]>
ref tikv#15279 Also, the timeout for etcd implementation was reduced to 5s. Signed-off-by: hillium <[email protected]> Signed-off-by: 山岚 <[email protected]>
close tikv#15340 Signed-off-by: Neil Shen <[email protected]> Co-authored-by: tonyxuqqi <[email protected]>
close tikv#15328 Signed-off-by: Yang Keao <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com>
…io. (tikv#15291) ref tikv#15070, close tikv#15268 Make a bugfix on the detection strategy on slowness when there exists io-hang exceptions. Signed-off-by: lucasliang <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com>
close tikv#15346 Make online unsafe recovery abort on timeout Signed-off-by: Connor1996 <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com>
…v#15307) close tikv#15306 add deadline check in get and batch_get request handler. Signed-off-by: crazycs520 <[email protected]>
close tikv#15344 Using space in log kv can cause slog to emit "", which inflates the log file. Also, we prefer underscores instead of hyphens as it aligns with Rust variable naming conventions. By unifying the style of kv names, makes it easier to search information during debugging. Signed-off-by: Neil Shen <[email protected]>
close tikv#15355 Disable `periodic-compaction` and `ttl` features as default. Signed-off-by: lucasliang <[email protected]>
close tikv#15316 Signed-off-by: glorv <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com>
ref tikv#15082 Record and print `duration to last update_safe_ts` and `duration to last consume leader`. Signed-off-by: ekexium <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com>
close tikv#15304 resource_metering: free leaked thread CPU stats Signed-off-by: Neil Shen <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com>
tikv#15385) close tikv#15384 Signed-off-by: Leavrth <[email protected]>
…#15342) ref tikv#15282 consider duplicated mvcc versions when check compact
close tikv#15388 Signed-off-by: wshwsh12 <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com>
…ket's version is stale (tikv#15224) close tikv#15123 reject cop request if the given buckets verison is stale, and return the new error called `bucket_version_not_match` which will carry with the latest bucket keys, so the client should update it cache depend on this errror. And then split this cop request again. Signed-off-by: bufferflies <[email protected]>
ref tikv#15401 Add some metrics about async snapshot Signed-off-by: SpadeA-Tang <[email protected]>
close tikv#15397 Signed-off-by: Yang Keao <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com>
close tikv#15371 Fix incorrect raftstore applys memory trace. When performing dropping apply fsm, it should reset the memory trace to zero instead of calling `update_memory_trace`, otherwise it would record extra footage of the interior vector capacity. Signed-off-by: Connor1996 <[email protected]> Co-authored-by: tonyxuqqi <[email protected]>
ref tikv#13311 Adding more log information for diagnosing the issue where inconsistencies between `region` and `region range` information in the meta may occur during region split and self-destroy scenarios. Signed-off-by: cfzjywxk <[email protected]>
ref tikv#15269 Make flow control of compaction pending bytes global in Partitioned-raft-kv as well. The major reason is because compaction pending bytes is a trend signal. If overall the TiKV's pending bytes is increasing, then it means the write speed is too fast and it needs to slow down. Essentially it means TiKV's capability is not enough to handle the traffic. And the side benefit is that it could greatly save the memory because we only need one VecDeq object for compaction pending bytes instead of one per region. The last reason to not make it regional is that the threshold is hard to set because the proper threshold depends on the region size (The idea is that if there're fewer regions, then we have higher tolerance in the extreme case if we only have one region, then it should be same as SingleRocksDB.), and that's the indicator that we should make it global. Signed-off-by: tonyxuqqi <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com> Co-authored-by: Neil Shen <[email protected]>
…s panel (tikv#15415) close tikv#15413 update procinfo to hanlde no data in context switch metrics panel Signed-off-by: SpadeA-Tang <[email protected]>
|
f622e28
to
6f36ad6
Compare
6f36ad6
to
fc22225
Compare
What problem does this PR solve?
Issue Number: close #xxx
Problem Summary:
What is changed and how it works?
Check List
Tests
Side effects
Documentation
Release note